An MRI-based radiomics-clinical nomogram for the overall survival prediction in patients with hypopharyngeal squamous cell carcinoma: a multi-cohort study

Objective To explore whether radiomics features extracted from pre-treatment magnetic resonance imaging (MRI) can predict the overall survival (OS) in patients with hypopharyngeal squamous cell carcinoma. Methods A total of 190 patients with hypopharyngeal squamous cell carcinoma were eligibly enrolled from two institutions. Radiomics features were extracted from contrast-enhanced axial T1-weighted (CE-T1WI) sequence. The least absolute shrinkage selection operator (LASSO) algorithm was applied to establish a radiomics score correlated with OS. Multivariate logistic regression analysis was applied to determine the independent risk factors, which was combined with radiomics score to build the final radiomics nomogram. Results A radiomics score with 6 CE-T1WI features for OS prediction was constructed and validated; its integration with specific clinicopathologic factors (N stage) showed a better prediction performance in the training, internal validation, and external validation cohorts (C-index 0.78, 0.75, and 0.75). Calibration curves determined a good agreement between the predicted and actual overall survival. Conclusions The radiomics-clinical nomogram and radiomics score might be non-invasive and reliable methods for the risk stratification in patients with hypopharyngeal squamous cell carcinoma. Key Points • An MRI-based radiomics model was constructed to evaluate of OS in patients with hypopharyngeal squamous cell carcinoma. • A radiomics-clinical nomogram that combined radiomics features and clinical characteristics was established. • Multi-cohort study validated the predictive performance of the radiomics-clinical nomogram to stratify patients with high risk in clinical practice.